基于3维矩阵变换的彩色图像质量评价方法研究

Quality Measure of Color Image Based on Three Dimensional Matrix Transform

鉴于传统的图像质量评价指标不能很好地反映图像质量,为了对图像质量进行更合理的评价,提出了一种新的彩色图像质量评价方法,该方法从图像的客观保真度、视觉失真度及清晰度3个方面来综合评价图像质量,其中,客观保真度作为图像质量的横向评价标准,而视觉失真度和清晰度则为其纵向评价标准。在评价图像的清晰度时,还充分考虑了彩色图像3帧间的结构相似性,把彩色图像的各个分量表示在一个统一的数学模型里,使用3维矩阵宽离散余弦变换来衡量图像的清晰度。实验结果表明,相对于传统的峰值信噪比评价指标来说,这种方法更为符合人眼视觉特性。

In this paper, DSF（ definition, error sensitivity and objective fidelity）, a new technique of color image quality measure has been proposed in which color image quality can he measured synthetically by calculating the objective fidelity, vision distortion and definition of a color image. The objective fidelity is a base for transverse comparison. While the vision distortion and definition is utilized for vertical evaluation . The great similarity among three frames of one image is taken into account fully and 3D-WDCT（ three dimensional wide DCT） is used when the definition was calculated. Three frames, R, G and B, are integrated into one single mathematics model to be transformed. Against PSNR, this method will provide a better performance consistent with human visual characteristic.